boltzgen

Other Community v1.0

Skill notes

When to Use This Skill

Use BoltzGen when you have: ✅ A cleaned target structure (CIF or PDB) with unambiguous chain IDs ✅ A defined epitope or hotspot set expressed in label_seq_id numbering ✅ A modality decision in hand — VHH single-domain, full antibody Fab, or de novo miniprotein ✅ GPU compute available — local (CUDA-capable, ≥24 GB VRAM), HPC (RunPod), or Tamarind cloud

Quick Start

A 20-design VHH preview against TNF-α (PDB 1TNF, epitope residues 45–52 and 78–85) on a local GPU: Expected runtime: ~10–25 min on a single H100/A100 (Stage 4 Protenix refolding dominates). Expected output: a CSV ranked by iptm descending, typically 60–80% of designs with iptm > 0.4, 20–40% with iptm > 0.6.

Installation

Software · Version · License · Commercial Use · Installation BoltzGen · ≥0.1.0 · MIT · ✅ Permitted · Clone https://github.com/HannesStark/boltzgen; install per repo README Protenix · v1.x (368M) · Apache-2.0 · ✅ Permitted · Bundled with BoltzGen install; downloads weights on first run AntiFold · ≥1.0 · MIT · ✅ Permitted · Bundled with BoltzGen install CUDA toolkit · ≥11.8 · NVIDIA EULA · ✅ Permitted · Match your driver version

Inputs

Required: Target structure file — CIF (preferred) or PDB Single chain or multi-chain Clean chain IDs (A, B, …) — no 0 or unusual letters Resolved coordinates for the epitope region (no missing density)

Outputs

All outputs land under the directory passed as --output. Primary results: File · Format · Description final_ranked_designs/final_designs_metrics_*.csv · CSV · Per-design metrics sorted by iptm descending final_ranked_designs/*.cif · CIF · Refolded structures for each ranked design

Fluxly profile Runtime and permissions are inferred by Fluxly static checks, not author declarations or runtime test results.
Domain
Other
Author-declared agents
No explicit declaration found; this is not inferred or tested compatibility
Static check
92 / 100 · heuristic scan, not runtime safety proof
Runtime evidence
Not run; no isolated runtime test evidence is currently recorded
Author / version / license
@001TMF · v1.0 · no license declared
Token usage
Moderate
Setup complexity
Manual integration
External API key
Required · Vendor-specific
Operating systems
Unspecified (assume cross-platform)
Runtime requirements
Python
Detected file/system behavior
  • Read-only
  • Write / modify
  • Shell exec
  • Env read
Network behavior
External requests
Install commands
None (reference only)

Heads up: Tool permissions are not declared; actual access depends on the runtime.

01
Step 1 — Write the entities YAML

Use the Write tool to create the spec file. The minimal form: Range notation rules and conversion tables: `references/entities-yaml-spec.md`. ✅ VERIFICATION: `yamllint workspace/design_spec.yaml` returns no errors.

02
Step 2 — Submit via boltzgensubmit.py

Targets: Target · Behaviour `local` (default) · Invokes `boltzgen run` directly on the current host `hpc` · Prints the deployment command and hands off to the by-deploy-compute skill `tamarind` · Submits via Tamarind API (requires `TAMARIND_API_KEY`)

03
Step 3 — Parse and rank

This walks `final_ranked_designs/`, handles multi-seed ensembles by merging all `final_designs_metrics_*.csv` files, and writes a single canonical CSV. ✅ VERIFICATION: Console shows `✓ parse_designs completed: <N> rows / <path>`.

04
Step 4 — Hand off to by-scoring / by-screening

The CSV is the input to downstream skills. Do not re-derive metrics inline. ⚠️ CRITICAL — DO NOT: ❌ Write inline `subprocess.run(["boltzgen", ...])` calls → use `boltzgen_submit.py` ❌ Parse the CSV with ad-hoc `pd.read_csv` filtering → use `parse_designs.py` then `by-scoring` ❌ Edit the entities YAML by hand on every run → use the script's `--binding-residues` flag to generate ranges

Discussion

Powered by GitHub Discussions. Sign in with GitHub to comment, react, or subscribe.